Abstract: Agent-oriented software engineering （AOSE） is viewed as a novel paradigm for complex systems. However, it is still a challenge to develop complex MASs that are dynamic, open and even self-adaptive with such technology. To deal with the problems, twofold should be considered when conducting the research on AOSE: the potentiality and flexibility of agent orientation paradigm should be extensively exploited and AOSE should borrow and integrate successful technologies and practices of software engineering. In this paper, an agent-oriented methodology called ODAM for adaptive MAS is presented, which is based on dynamic binding mechanism, borrows organization metaphor to model adaptive MAS in an abstract and natural way, and integrates iteration development and MDA approach to simplify the development and adapt to the variety of agent technologies and platforms. The methodology framework and technical details are introduced, including dynamic binding mechanism, meta-model and modeling language based on organization abstractions, and software development process based on iteration development and MDA. A case is also studied to illustrate the approach.